(复杂系统的性能评价与化课件资料）LN ProbQueue.docVIP

Lecture Notes #2 Probability and Queueing Theory (Version 3.5, 2007-9-17) (See also [CL99] §6.7, 8.2, and 8.5, Chapter 10.1-10.7) 1. Probability Fundamentals 1.1 The memoryless triangle The single most important fact to commit to memory: The memoryless triangle: (i) The inter-arrival time distribution of the customers is exponential. Let T be an exponentially distributed random variable with mean 1/?, i.e., p(T)=?e-?T. (We use capital P to indicate the probability, and small p to indicate the probability density function.) Then the conditional probability of T, given that T is greater than any finite value t0, is the same as the original distribution, i.e., P(T≤ t+to | Tto) = 1- e -?t = P(T≤t). (1) ( p(t) = ?e-?t The intuitive explanation of equation (1) is: P(at least one arrival in the next t seconds | no arrival in t0 seconds) = 1-P(no arrival in t seconds) = P(at least one arrival in t seconds). This property, illustrated in Fig. 1, is called the memoryless property. The shaded portion of the unconditional density function is re-scaled to give the conditional distribution Fig. 1 The Memoryless Property of the Exponential Distribution To prove (1) consider (ii) The probability of a single arrival (coin toss, head or tails) in time ?t is equal to ??t + o(?t), where ???is the so-called constant arrival rate (??t is the probability of head), and the arrivals are independent. Thus the probability of no arrival in [0,t] (i.e., the arrival instant ta is greater than t) can be obtained by dividing the interval into N equal increments of ?t = t/N, and which implies (i). In the above equation, (1-??t) in the second term means P(no arrival in ?t), and (1-??t)N means P(no arrival in all N intervals). From this equation, we also have and p(ta)=p(T)=?e-?T. For finite and small ?t, the distribution is known as geometric distribution. (coin toss explanation) (iii) The probability of n arrivals in the interval [0,t] obeys the Poisson distribution. Let